Novel network proposed to improve underwater image quality
Where do the researchers work?
The researchers work at the Hefei Institutes of Physical Science of the Chinese Academy of Sciences. They are led by Prof. Wang Rujing and Wang Liusan.
What journal published the research results?
The research results were published in Frontiers in Marine Science. The study focused on addressing the problem of underwater image quality degradation by establishing a learnable full-frequency transformer dual generative adversarial network (LFT-DGAN).
What does LFT-DGAN stand for?
LFT-DGAN stands for Learnable Full-Frequency Transformer Dual Generative Adversarial Network. It is a novel approach proposed by researchers to enhance the quality of underwater images. The method uses a combination of reversible convolution-based image decomposition, a robust dual-domain discriminator, and a transformer model to effectively separate and extract features from different frequency domains, addressing challenges unique to underwater applications4.